时 间：2024年10月30日16:00 -17:00
地 点：普陀校区理科大楼A302
报告人：黄译旻 诺亚方舟实验室研究员
主持人：徐进 华东师范大学教授
摘 要：
In recent years, many industries start using AI algorithms to help their business. First of all, Black Box Optimization problems are everywhere in their scenarios. As a very effective global optimization algorithm, Bayesian Optimization uses advanced probabilistic surrogate model and the acquisition function appropriately to guarantee to obtain the optimal solution under a few numbers of function evaluations. And it also doesn’t need any information of the objective function, in which their objective functions could not be expressed, or the functions are non-convex, and computational expensive. Thus Bayesian Optimization is very suitable to solve Black Box. This talk will analyze several research directions in BO, and introduce our achievement both in statistics and industry application.
报告人简介：
黄译旻，诺亚方舟实验室研究员。北京大学数学科学学院概率统计系本博，试验设计方向。2019年加入诺亚，将统计理论应用于机器学习，聚焦黑盒优化及数学规划求解器方向研究。在MLSys, NeurIPS, ECCV, AAAI, JSPI等期刊会议上发表多篇相关工作。